Even if $I+J=(1)$ for two-sided ideals, we don't necessarily have $I \cap J = IJ$
sums of nilpotent elements need not be nilpotent. In particular, the set of all nilpotent elements is not an ideal
over a division ring, polynomials in one variable can have infinitely many roots
When A is a ring, then and $a \in A$, then the evaluation map $A[x] \to A, f \mapsto f(a)$ doesn't have to be a ring homomorphism (it works iff $a$ is central, basically because $x$ commutes with every element of $A$ inside $A[x]$)
The question was to give an example of a polynomial ring $R[x]$ over a base ring $R$ and polynomials $h(x), f(x), g(x)$ such that $h(x) = f(x)g(x)$ but $h(a) \neq f(a)g(a)$ for some $a \in R$
What you said makes sense because $R[x]$ is commutative (always?) but $R$ need not be
technically, one also needs to specify that we evaluate with coefficients on the left since that's not the same as evaluating with coeffients on the right
@Daminark do you want to see some weird noncommutative ring?
Let $k$ be a field and $V$ be an infinite dimensional vector space over $k$. Note that we have $V \cong V \oplus V$, fix such an isomorphism $f: V \to V \oplus V$ and denote the components of $f$ by $f_1$ and $f_2$. Let $R=\operatorname{End}_R(V)$ (basically infinite matrices)
If we have $\varphi \in R$, we can compose with $f_1$ and $f_2$ from the right and this gives us again $k$-linear maps $V \to V$, i.e. elements of $R$. So we get a map $R \to R^2$ given by $\varphi \mapsto (\varphi \circ f_1, \varphi \circ f_2)$
Also side note, the functional analysis book that I recommended to GFauxPas earlier is one you might like (since you mentioned you didn't like your functional class doing more numerical stuff). Functional Analysis, Spectral Theory, and Applications
Our professor probably would've used it for our class if he knew about it then
it's easier if you think in Hom-Sets. $R = Hom_k(V,V) \cong Hom_k(V,V \oplus V) \cong Hom_k(V,V) \oplus Hom_k(V,V) = R^2$ and since we only changed things in the second component, if we compose (i.e. multiply) we can pull that out, so it's left $R$-linear
I wanted to write down the maps explicitly but I messed up somehow